fix: post-merge review fixes for v0.2.10 by rejojer · Pull Request #402 · VectifyAI/PageIndex

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer · 2026-08-13T06:53:47Z

Summary

Conformant responses() envelope: official output (model items only) + items (full transcript for round-trip), verified against real OpenAI API
Usage details aggregated across turns (cached_tokens, reasoning_tokens) on both OpenAI surfaces
Python floor corrected to >=3.10 (litellm stable channel requires it; verified in clean venv)
Three stale anthropic>=0.84.0 hint strings updated to 0.108.0
Image-stub behavior disclosed on as_openai_tools / as_anthropic_tools docstrings
Non-essential comments trimmed (−61 lines)

Test plan

267 tests passed locally
Live A/B vs real OpenAI API: field-identical to official hand-rolled flow
Response.model_validate passes on non-stream envelope
Stream events validate against official ResponseStreamEvent union
Clean 3.10 venv install + full suite verified
Round-trip prefix tests confirm cache contract unchanged

… items, full usage details - output now carries only model-produced items, so the envelope parses with the official openai SDK types (function_call_output is input vocabulary — the real API never returns it in output) - the full process transcript moves to the new items field; round-trip appends items instead of output (same bytes, so the provider prompt-cache prefix contract is unchanged) - usage aggregates token details across turns (cached_tokens, cache_write_tokens, reasoning_tokens) on both OpenAI surfaces — cache hits are now observable instead of discarded - streaming stops synthesizing the nonstandard tool-output event; every stream event now validates against the official event union, tool results arrive in the terminal envelope's items - tests: two conformance tests pin the contract (non-stream model_validate + per-event stream validation); round-trip prefix tests append items Verified: 267 tests green; live A/B against the real OpenAI API — field-identical to the official hand-rolled flow, round-trip accepted with zero repeat tool calls.

pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

…ertising tool surfaces as_openai_tools / as_anthropic_tools cloud docstrings advertised the image tool without mentioning that the in-process bridge replaces base64 payloads with text placeholder stubs (mcp_bridge call_tool).

litellm's stable channel (every release satisfying our >=1.84.0 floor) and both agent extras require 3.10; on 3.9 pip resolution fails on the hard deps (verified in a clean venv — zero packages install). A clean 3.10 venv with all three extras runs the full suite green. CI already tests 3.10/3.13 only. The >=3.7 claim was inherited from the two-dep 0.2.8 client and was already unsatisfiable then (openai>=1.70 needs 3.8). Closes recurring review finding #10.

53 comment lines removed: rationale that belongs in commit messages, descriptions restating what adjacent code or function names already show, and cloud-implementation provenance notes. Section headers and constraint comments (protocol invariants, safety guards) kept.

rejojer added 7 commits August 13, 2026 14:55

fix: stale anthropic>=0.84.0 hints — the supported floor is 0.108.0
62f8f28
pyproject raised the floor in f58cca1 (0.84-0.107 execute a refusal turn's tool_use blocks); the three user-facing strings still pointed hand-installers at the broken range.

refactor: trim the envelope-change comments to the essential constraint
5ee2749

chore: trim rationale comments from this session's commits
3c65db5

rejojer force-pushed the fix/post-merge-review branch from 5860c7c to 9485de6Compare August 13, 2026 06:56